Development of Economical Analysis and Technical Solutions for Efficient
Distribution Transformers
A.F. Picanço1*, C de Salles1, M.L.B. Martinez1, P.C. Rosa1, H.R.P.M. de Oliveira2
1
Federal University of Itajubá and 2AES Sul Utility Company
Rua Cel. Rennó, 05, Itajubá, MG, Brazil, CEP 37500-050
Telephone number: +55 35 3622 3546 and E-mail address: alessandra@lat-efei.org.br
ABSTRACT: Transformers are equipments applied to
electrical systems to adjust voltage levels for consumption,
distribution, transmission, and generation needs. Actual
technology indicates that they are responsible for
approximately one third of total network losses. These
losses can be accounted as power utility costs, costs to
society and to the environment itself. The aim of this paper
is to study the relationship between the transformer costs
and its losses. A method for collecting cost data and for
building cost surfaces - the so called “Production and Total
Cost Surfaces” is presented. The economic analysis of
transformer designs considers a variation of the reference
data of the main constructive parameters of a standard
transformer. Therefore, changes are introduced, amongst
others, in the core dimensions, such as the column
circumscribed area, its diameter and the yoke area. This
approach helps the power utilities to purchase equipment
according to the forecast demand, decreasing the total
network losses.
Keywords: Cost of Losses, Demand, Design, Distribution
Transformer, Energy, Optimization.
I. INTRODUCTION
Operational power losses of distribution transformers are
basically related to the core – no-load losses and to the
windings – load losses. The load losses depend on the
square of the load current during a standard daily operation
cycle. The no-load losses depend on the core magnetization
and can usually be considered as being constant during
standard operation.
The losses are an important parameter in the evaluation
of the total transformer costs i.e. investment and operation
costs. They can lead to a reduction of the utility net
economical result and can, ultimately, be considered as a
“Loss to Society”.
In distribution systems, transformers are responsible for
roughly one third of the total power losses. This is the
driving force behind the loss and cost reduction efforts [1].
The aim of the harmonization of the European Union (EU)
documents (HD 428 and HD538) and of the United States
(US) NEMA TP-1 and Technical Support Document
(TSD) [2] is the reduction of the loss costs and
consequently the decrease of power losses in the network.
These documents evaluate the capitalized costs and the
efficiency of transformers considering their daily loads.
The first step in designing an efficient transformer for a
specific region is to carefully model the supplied load. The
parameters that define the load behavior are basically the
load, peak and factor, and the load increase rate.
Some utilities in Brazil normally purchase stacked core
transformers that are built using steel silicon grade M2 or
M3 in rectangular geometry. Therefore, low density
magnetic flux is obtained increasing the transversal section
and results in reduced no-load losses. The windings are
normally constructed using copper wiring or foils and in
this case the increase of the mass of copper results in
reduced load losses. This also results in an increase of the
core window and also in the no-load losses [3-6]
The proposal of this study is to solve this contradiction
through the analysis a set of transformer designs presenting
reduced losses. This subject has been object of study since
the 60´s [4-8]. The solution is to search out an optimal
transformer design, considering the restrictions imposed by
the short circuit impedance, the daily load, the interest rate
and the energy price.
The present study is divided, firstly in the recognition of
the problem, secondly in the construction of a set of
problem solutions and thirdly, in the search for the optimal
design.
II. MANUFACTURING AND TOTAL COST
SURFACES
The manufacturing and total cost surfaces show a set of
design possibilities considering manufacturing cost and
total cost versus no-load and load losses. This method
indicates the interaction between investment and
operational costs and losses.
The manufacturing costs consider the quantity and the
price of the commodities used for building the transformer,
as defined by (1). [9,10].
C Manufacturing = C fixed + Ccore M core + Ccopper M copper (1)
Where:
CManufacturing is the manufacturing cost in US$;
Cfixed is the transformer fixed cost in US$;
Ccore is the core material cost in US$;
Ccopper is the cost of copper in US$;
Mcore is the core total mass in kg;
Mcopper is the copper total mass in coils in kg.
The total, or operational, cost of a distribution
transformer is obtained by adding the manufacturing costs
to the no-load loss costs, and to the load losses considering
a daily loading cycle, as shown by (2). [3].
C Total = C TR / Analysis + CW 0 + CWL (2)
Where:
CW0 is the cost of no-load losses in US$;
CWL is the cost of load losses in US$;
CTotal is the total cost of transformer in US$;
 CTR/Analysis is the transformer cost in US$.
The first step of this method considers variations in
design parameters: magnetic induction density – no-load
losses, LV winding current density and HV winding
current density – load losses. Each point on the
manufacturing surfaces - Figure 1- represents a transformer
design.
Fig. 1 – Manufacturing cost surfaces considering variations in LV
winding current density, HV winding current density and magnetic
induction density.
The total cost depends on the impact of the energy costs
on the no-load and load losses. The load loss costs can be
calculated based on a parameter defined as “Time
supplying the maximum rated power”, TSMP, (3) or the
period of the day, in hours, that the transformer operates in
full load condition with the same area – “energy” – below
the load cycle profile.
Figure 2 shows the load cycle used to calculate the total
costs in this paper. The TSMP is 8.72.
2
⎛ kVAi
24
⎞
TSMP = ∑ ⎜⎜ ⎟⎟
i =1 ⎝ kVAN ⎠
Fig. 2 – Load cycle used as an example to generate the total cost
surface presented in Figure 3
The behavior of the total cost surfaces versus losses for
the load cycle of Figure 2 is shown in Figure 3.
III. DESIGN SURFACES: ESTABLISHING A SET OF
SOLUTIONS
The second step of this method is to define sets of
solutions. The searching tool will try to find an optimal
design in these sets.
Three-dimensional matrixes are built considering the
following design parameters:
a) thickness of the conductor of the LV winding;
b) width of the conductor of the LV winding;
c) diameter of the conductor of the HV winding;
d) magnetic induction;
e) insulation thickness;
f) clamp dimensions;
g) design of the end insulation;
h) Gap between LV and HV windings.
Fig. 3 – Total cost surfaces considering variations in LV winding
current density, HV winding current density and magnetic induction
density.
This method considers design possibilities stored in
three-dimensional matrixes (11x11x22), as in Figure 4, and
design surfaces as in Figure 1.
(3)
Fig. 4 – Scheme of three-dimensional 11x11x22 matrix
The design surfaces allow verifying the influence of each
constructive parameter on the losses and the associated
manufacturing cost impact. Considering the current density
of the HV winding and magnetic induction as being
constant, it is possible to verify that the increase of the
thickness of LV conductor results in two opposite effects;
the decrease of the load losses and the increase the no-load
losses, effects that must be properly balanced. It is also
observed that the short circuit impedance is acceptable only
for a determined thickness range of the LV conductors.
For this specific transformer design, the behavior of
short circuit impedance is as shown in Figure 5, where
variations on the current density of the HV winding, on the
insulation thickness and on the magnetic induction were
considered. The shape of this trend is not standard once it
depends on the transformer rated power, rate voltages and
supplied load. Therefore each specific case must be
properly studied and analyzed.

Fig 5 – Short circuit impedance surface versus losses
Figure 6 presents the hot-spot temperature surfaces
versus short circuit impedance and payback in years. The
lowest hot spot temperature [11], which is associated to the
transformer life span, is about 137°C. The design has a
current density of the HV winding, an insulation thickness
and a magnetic induction 50% higher compared to a
standard transformer. This, as occurs in other cases, is not
feasible, due to manufacturing and cost restrictions. Some
of the cost restrictions are well defined by the negative
payback results that are associated to the solution process
of the cost equations.
Fig 6 – Hot-spot temperature surface versus short circuit impedance
and payback.
The design surfaces presented in this paper are based on
solution sets considering variations on several transformer
parameters and grouped in the following way:
a) Variations of current density of the LV and HV
windings and magnetic induction;
b) Variations of current density of the HV winding,
insulation thickness and magnetic induction;
c) Variations of current density of the HV and LV
winding, insulation thickness;
d) Variations of current density of the HV winding, gap
between coils and magnetic induction;
e) Variations of the design of the end insulation,
clamping dimensions, insulation thickness and
magnetic induction;
f) Variation of the gap between coils, insulation
thickness and magnetic induction;
g) Variation of current density of the LV winding,
insulation thickness and design of the end insulation
and clamp dimensions;
h) Variation of current density of the HV winding,
insulation thickness and gap between coils.
IV. OPTIMAL DESIGN BASEAD ON THE SURFACES
In the present study, the optimal transformer design must
present reduced loss costs. Therefore, an optimization tool
[12, 13] is applied to the design solution sets to find a
minimum cost point on a surface.
The points on the total cost surfaces represent an element
of the three-dimensional design matrixes, shown in Figure
4. The comparative and iterative procedure adopted to
search for the minimum cost point is as follows:
a) The first minimum cost is a random value;
b) For the iteration k, in the column j, each element
c(i,j,k) is compared with the minimum cost value;
[c(1, j, k ), c(2, j, k ),K, c(N , j, k )]T
c(i, j, k ) < C Mínimo
c) The search must satisfy the level of the no-load losses,
load losses and short circuit impedance according the
national or utility standards in the set of solutions;
WN (i, j , k ) ≤ W N ( ABNT )
WTotal (i, j , k ) ≤ WTotal ( ABNT )
Z %(i, j , k ) ≤ Z % ( ABNT )
d) The payback is defined by the customer.
0 < TRI (i, j, k ) ≤ 5 years
In some cases, the local minimum point results in a
design in which the purchase price is not attractive to the
customer. Because of this, global minimum points located
between the local minimum and the costs presented by a
standard transformer can be attractive economical
solutions.
The search procedure and determination of the global
points is as follows:
a) Find the factor R: the ratio between the minimum
and the standard transformer total costs;
b) If R>1, the global point is close to the standard
transformer total cost;
c) If R<1, for each iteration, a multiplier M(k) is
defined as follows:
M (k ) = R + 0,1 ⋅ k
d) If M(k)<1, a value for the minimum total cost
CMin(k) is estimated. This point is close to the local
minimum tending to move toward the standard
transformer total cost. In this way, all the global
minimum points are collected;
C Min (k ) = C S tan dard ⋅ M (k )
e) For each iteration k, and for each column j, the total
cost element c(i,j,z) is compared to the estimated
minimum total cost;
C Min (k − 1) < c(i, j , z ) < C Min (k )
Figure 7 presents the flow chart of the above discussed
process.
V. OPTIMAL DESIGN RELATED TO THE TIME
SUPPLYING THE MAXIMUM RATED POWER
In this paper, the economic restrictions for the
application of the proposed optimization process are: an
energy cost of 59.87 US$/MWh, an interest rate of 8% per
year, and a 10-year analysis period.
presents one local (k=0) and another ten global minimum
points.
Table 1 shows a comparison between the technical
characteristics of the efficient and standard designs. Table 2
presents a comparison between the payback and operational
cost reduction and energy saved.
This analysis considers point (k=0) as being the “best”
solution because it presents the maximum total cost
reduction. However, from a payback perspective, this is not
the “best” solution. Assuming that all design options in
Table 1 are feasible for manufacturing, the final solution
will be related to the utility policy. For instance, from a
payback standpoint, the answer is design 7. However,
design 9, which presents the second operational cost
reduction, would also be a possible choice.

Table 2 – Economic results attached to Table 1

Reduction Operational
Pay-
of the Cost Reduction Energy Saved
k back,
Total Cost per unit, [MWh/Year]
[Years]
[%] [US$/Year]
0 5,66 2,50 75,09 1,3248
1 5,60 2,06 67,02 1,1195
2 5,40 2,20 66,07 1,1036
3 5,35 2,22 65,82 1,0995
4 4,91 2,49 64,40 1,0757
5 4,84 3,00 72,39 1,2092
6 4,78 3,26 59,84 0,9996
7 4,67 1,98 54,29 0,9068
8 4,66 2,02 54,36 0,9125
9 4,34 3,31 70,64 1,1801
10 4,04 2,38 51,51 0,8604

Fig 7 – Flow chart to find global minimum points.

Figures 8 to 10 show three daily load cycles, presenting

different load peaks and TSMP. With them it is possible to
express the concept that for each TSMP there are several
efficient transformers to supply the load. Some of them
have the “best” manufacturing characteristic and therefore
are a logical choice. In theory this is the “optimal technical
and economical solution”: One efficient transformer per
load. However, spare, maintenance and purchase polices of
the utility, together with mass manufacturing restrictions
can modify this and, in the end the final design option may
point to a “unique” solution, with an improved economical Fig 8 – Load characteristic for a TSMP of 12.56
profile.
The minimum total cost (k=0) is a design that presents
Table 1 – Transformer designs presenting losses lower than the standard
Reduction of Reduction of No- Reduction of
reductions of 50% on the current density of the HV
k winding, 30% on the gap between the windings and 40%
Total Cost [%] Load Losses [%] Load Losses [%]
0 5,66 4,44 33,60 on the dimensions of the end insulation. The hot-spot
1 5,60 3,15 30,36 temperature is a common concern for the utility.
2 5,40 2,70 30,11 Considering a daily ambient temperature of 40°C and the
3 5,35 3,03 29,84 load cycle of Figure 8, during the period from 6 to 7 p.m.,
4 4,91 1,46 29,88 the top-oil and the hot-spot temperature are around 107°C
5 4,84 2,01 33,42
and 140°C, respectively, a reduction of approximately 1°C
6 4,78 0,64 28,10
7 4,67 1,93 24,87
compared to the design of a standard transformer [4].
8 4,66 0,90 25,50 Figure 9 shows a daily load cycle with a TSMP of 8.7.
9 4,34 2,17 32,52
10 4,04 1,66 23,68

For a daily load cycle according to Figure 8, having a

TSMP of 12.56, and considering variations on the current
density of the HV winding, on the gap between the coils
and on the design of the end insulation, the set of results
 Fig 9 – Load characteristic for a TSMP of 8.7 Fig 10 – Load characteristic for a TSMP of 4.5

This case considered variations on the current density of This new case considered variations on the current
the HV windings, on the insulation thickness and on the density of the HV windings, on the insulation thickness and
gap between the LV and HV windings, a different solution on the gap between the LV and HV windings. Tables 5 and
compared with the previous one. Tables 3 and 4 show a 6 show a comparison between the technical and economical
comparison of the technical and economical characteristics characteristics presented by the efficient and standard
presented by the efficient and standard designs. The set of designs. The set of results presents one local and six global
results presents one local and seven global minimum minimum points.
points. The three studies presented have a different number of
global minimum points. This is a normal result and is
Table 3 – Transformer designs presenting losses lower than the standard related to the optimization process and to some technical
Reduction of Reduction of No- Reduction of Load restrictions.
k
Total Cost [%] Load Losses [%] Losses [%]
0 5,57 2,92 36,25 Table 5 – Transformer designs presenting losses lower than the standard
1 5,53 2,82 36,23
Reduction of Reduction of No- Reduction of
2 5,14 2,04 35,88 k
Total Cost [%] Load Losses [%] Load Losses [%]
3 5,13 3,18 23,47
0 3,14 3,18 23,47
4 4,97 2,77 23,88
1 2,98 2,77 23,38
5 4,43 1,86 22,58
2 2,46 1,86 22,58
6 4,15 0,31 29,12
3 2,24 2,92 36,25
7 3,35 0 20,98
4 2,20 2,82 36,23
5 1,81 2,04 35,88
Table 4 – Economic results of Table 3 6 1,44 0,31 29,12
Operational
Reduction of Pay-
Cost Reduction Energy Saved
k the Total back, Table 6 – Economic results of Table 5
per unit, [MWh/Year]
Cost [%] [Years] Operational
[US$/Year] Reduction
Pay- Cost
0 5,57 1,71 55,85 0,9329 of the Energy Saved
k back, Reduction
1 5,53 1,73 55,72 0,9308 Total Cost [MWh/Year]
[Years] per unit,
2 5,14 1,95 54,48 0,9100 [%]
[US$/Year]
3 5,13 0,09 37,39 0,6246 0 3,14 0,16 20,92 0,3495
4 4,97 0,18 36,87 0,6149 1 2,98 0,33 20,08 0,3417
5 4,43 0,50 34,83 0,5818 2 2,46 0,93 18,98 0,3171
6 4,15 1,85 42,93 0,7172 3 2,24 3,33 30,41 0,5080
7 3,35 1,28 30,75 0,5137 4 2,20 3,39 30,30 0,5061
5 1,81 3,91 29,29 0,4893
In this case, one recommended solution is design (k=2), 6 1,44 3,81 22,49 0,3757
which presents a reduction of 30% in the current density of
the HV winding, 10% of insulation thickness and 20% of Design 2 (k=2) presents a good solution, in that it
the gap between the LV and HV windings when compared presents a reduction of 10% on the current density of the
to the standard project of a distribution transformer. This HV winding, 10% of the insulation thickness, which must
can be attractive due to the fact that the technical and be verified by proper dielectric checking, and a rise of 20%
economical characteristics are quite close and that this of the gap between the LV and HV windings as compared
solution presents a lower manufacturing cost compared to to the standard values. In this case the amount of copper is
the local minimum point (k=0), reducing the investment. reduced and the solution presents an operational cost close
Figure 10 shows a daily load cycle with a TSMP of 4.5. to the minimum point. The payback is 0.93 years and there
is an energy savings of 0.3171 MWh per year.
Design 5 (k=5) is also a satisfactory solution when
energy savings is a priority in the utility policy. The
constructive parameters result in a reduction of 30% on the
current density of the HV winding, 10% of the insulation
thickness and 20% of the gap between the LV and HV
windings. The payback is 3.91 years and energy saved is
0.4893 MWh per year.
Finally, the shape of the total cost surface for this case is
shown in Figure 11.
7800
7600
7400
Total Cost
7200
7000
6800
6600
2000
185
1000 180
0 175
Wcu Watts W0 Watts
Fig 11 – Surface for the set of designs with variation on the current
density of HV winding, on the insulation thickness and on the gap
between LV and HV windings considering TSMP of 4.5
VI. CONCLUSION
The time supplying maximum rated power (TSMP) is the
first parameter to be considered in the design of an efficient
distribution transformer. This is related to the influence this
parameter bears on the cost of the transformer designs,
mainly because it directly controls the load losses and plays
an important indirect role in controlling the no-load losses.
A statistical study of the load is necessary to define one
or several load profiles. The spare and maintenance policies
of the utility also play an important role in the process of
loss cost reduction. Therefore, these issues, together with
transformer sampling to define design loads, are key factors
to reach optimal solutions and must be object of careful
consideration.
As shown, each load profile presents a set of efficient
transformers. This is not a practical solution for a utility;
however, this can be useful for industrial consumers. The
final choice must consider financial restrictions,
manufacturing process restrictions and recommendations,
the price of the main transformer commodities, the utility
internal policy and governmental energy saving
regulations, all of which are normally conflicting issues and
therefore must be properly balanced.
VII. ACKNOWLEDGEMENT
Alessandra Picanço acknowledges the financial support
of CAPES, AES-Sul S.A. and PROCOBRE BRASIL. She
also acknowledges the Prof. Dr. Estacio Tavares
Wanderley Neto for your support.
VIII. REFERENCES
[1] European Copper Institute. “The Scope for Energy
Saving in the EU through the Use of energy-Efficient
Electricity Distribution Transformers”. European
Communities, 1999.
[2] U.S. Department of Energy. “Distribution
Transformers Final Rule Technical Support Document –
TSD”. Washington, DC 20585, 2007.
[3] Blume, L.F., Boyajian, A., Camilli, G., Lennox, T.C.,
Minneci, S., Montsinger, V.M. “Transformer Engineering:
A Treatise on the Theory, Operation, and Application of
Transformers”. John Wiley&Sons, Inc., 1951.
[4] Bean, R.L., Chackan, N., Moore, H.R., Wentz, E.C.,
“Transformers for the Electric Power Industry”. McGraw-
Hill Book Company, Inc., 1959.
[5] DelVecchio, R.M.,Poulin,B.,Feghali, P.T.,Shah,
D.M., Ahuja, R. “Transformer Design Principles” CRC
Press, 2001.
[6] Kulkarni, S.V. and Khaparde, S.A., “Transformer
Engineering: Design and Practice”. Marcel Dekker, CRC
Press, Taylor&Francis Group, 2004.
[7] Putman, T.H. “Economics and Power Transformer
Design” IEEE, December 1963.
[8] Jabr, R.A., “Application of Geometric Programming
to Transformer Design”, IEEE Transaction on Magnetics,
Vol. 41, No.11, November 2005.
[9] Picanço, A.F., de Salles, C., Rosa, P.C., Martinez,
M.L.B., Oliveira, H.R.P.M., de Jesus, N.C. “Designing
Transformers According to Load”. 15th ISH International
Symposium on High Voltage Engineering, Ljubljana,
Slovenia, August 2007.
[10] Targosz, R. “The Potencial for Global Energy
Savings from High Efficiency Distribution Transformers”.
Leonardo Energy Transformer – European Copper
Institute, February, 2005.
[11] IEEE C57.91 – 1995. “IEEE Guide Loading
Mineral-Oil-Immersed Transformer” IEEE Power
Engineering Society, 2003.
[12] Fletcher, R. “Practical Methods of Optimization,
John Wiley & Sons, 1980.
[13] Pierre, A.D. “Optimization Theory with
Applications”. Dover Publications, Inc., 1986.
For reading:
[1] Picanço, A.F. “Avaliação Econômica de
Transformadores de Distribuição com Base no
Carregamento e Eficiência Energética”. Thesis of Master
Degree, Federal University of Itajubá, Brazil, April 2006.
IX. BIOGRAPHIES
Alessandra Freitas Picanço was born in 1978 in Manaus, Amazonas
State, Brazil. She received her B.Sc. in electrical engineering in May 2002
from the Federal University of Amazonas – UFAM and M.Sc. degree in
May 2006 from Federal University of Itajubá – UNIFEI. She is currently
obtaining her Doctor Degree in Transformer Designs.
Credson de Salles was born in 1975 in Bueno Brandão, Minas Gerais
State, Brazil. He received his B.Sc and M.Sc in electrical engineering in
from the Federal University of Itajubá where nowadays is a PhD Student.
Hermes R. P. M. de Oliveira was born in 1960 in São Lourenço, Minas
Gerais State, Brazil. He received his B.Sc. and M.Sc. degrees in electrical
engineering from EFEI (UNIFEI). He has worked in AES Sul Company
since 1998.
Manuel L.B. Martinez was born in 1959 in Santos, São Paulo State,
Brazil. He received his B.Sc. and M.Sc degrees in electrical engineering
from the Federal University of Itajubá (UNIFEI) and Ph.D. degree from
PEA-USP. He has been responsible for the High Voltage Laboratory of
UNIFEI since 1995.